ADS1120-Q1: About the noise performance(VRMS/VPP)

Hi Tao2199,

See my responses below.

Best regards,

Bob B

Tao_2199 said:

Part Number: ADS1120-Q1

Hello,

 

Regarding to the noise performance(table.1, page61) on ADS1120-Q1,my customer is asking some question.

 

(Question)

(1) I’m understanding about the conversion from noise VPP to noise VRMS by following formula.

VRMS = (VPP/2) /√2

But almost noise VRMS value are same. Why are these data are same?

(The calculation does not match by above formula.)

Could you please tell me this reason? [Bob] First off the calculation given above relates to a sine wave and not the Gaussian noise distribution.  Generally noise Vpp relates to 2 times the crest factor (usually 3 or 3.3) times the Vrms (Vpp = 2 * CF Vrms).  With delta-sigma converters the noise is measured with shorted inputs and either with a specific number of samples or a specific period of time.  For the ADS1120 a specific period of time is used to make sure drift is not a part of the noise measurement.  The measured results may not directly correlate with the calculated. Often times the measured Vpp is better than the calculated and that is what is in the datasheet.  As far as the noise Vrms and Vpp being the same value, the ADS1120 is the 16-bit version of the ADS1220 which is 24-bit.  The quantization noise is less than 1 code so the certainty is 1 code, but measured you see no change between the Vrms and Vpp.

 

(2) According to 7.1 Noise Performance(page61), the input referred noise drops when reducing the output data rate.

But the almost input referred noise(VRMS) in figure.1 is same during all data rate.

Why are these data same during this data rate range?

Could you please tell me this reason? [Bob]  I think you meant table 1 and not figure 1.  I think I have answered this in the first response.  If you look at a gain of 128 both in table 1 and table 2 on page 15 of the ADS1120-Q1 datasheet you can see that one code is about 490nV, so the standard deviation (as rms or effective bits) stays within one code from 330 sps and lower while the Vpp or noise-free bits is fractional (except for gain of 1).  What the table results come down to is if the noise cannot be determined because it is less than one code, you will not see significant changes.

 

(3)However, only input referred noise values(VRMS) during particular rate at gain=64-128 are reduced by the output data rate.

Why are these data reducing during only this data rate range/gain? [Bob] As I previously stated you first need to use the correct calculation and you need to remember that measurable capability of the device.

 

Regards,

Tao2199